PDA

View Full Version : Can the GeForce FX be "fixed" with drivers?


Pages : [1] 2 3 4 5 6 7

SurfMonkey
01-31-03, 07:39 AM
Carmack was told that there was plenty of room for improvement in the drivers yet. Will this make a big difference?

Hanners
01-31-03, 07:49 AM
Looking at some of the numbers in benchmarks, there has to be some room for improvement. The Pixel Shader 2 test numbers for example are horribly low compared to what I would expect.

As for exactly how much - Who knows?

A lot of the most important situations where the GeForce FX loses out seem to be down to it's hardware limitations though, rather than drivers. I don't think performance with AA/AF is going to improve.

Smokey
01-31-03, 07:59 AM
Honestly, should this question really be asked? Anyone who thinks that the drivers are not going to improve, and improve they will, I think are people who dont want them to improve! Anyone who has ever owned an Nvidia cards, knows 100% they get looked after when it comes to drivers, things are not going to change :rolleyes:

Hanners
01-31-03, 08:04 AM
Originally posted by Smokey
Honestly, should this question really be asked? Anyone who thinks that the drivers are not going to improve, and improve they will, I think are people who dont want them to improve! Anyone who has ever owned an Nvidia cards, knows 100% they get looked after when it comes to drivers, things are not going to change :rolleyes:

The question is not whether they are going to improve (of course they will), but whether they will improve performance.

YeuEmMaiMai
01-31-03, 05:03 PM
i definately know that unless there are some serious hardware flaws, Nvidia is more than capable of improving the drivers by a good margin.


question is will it be enough to beat Ati? I do not think so overall

Smokey
01-31-03, 06:45 PM
Originally posted by YeuEmMaiMai
i definately know that unless there are some serious hardware flaws, Nvidia is more than capable of improving the drivers by a good margin.


question is will it be enough to beat Ati? I do not think so overall

Your funny :rolleyes:
"Nvidia is more than capable of improving the drivers by a good margin"

Then you said

"will it be enough to beat Ati? I do not think so "

When Nvidia improves the drivers, and by a good margin, the GF-FX will be ahead in every benchmark. So I have to ask, what do you mean by "will it be enough to beat Ati? I do not think so " :wtf:

Shinri Hikari
01-31-03, 07:02 PM
If that is what you believe, more power to you. I am a Maybe...:|
I am however still going to buy it.:thumbsup: :cool2: :jumping:

Steppy
01-31-03, 07:33 PM
Originally posted by Smokey
Your funny :rolleyes:
"Nvidia is more than capable of improving the drivers by a good margin"

Then you said

"will it be enough to beat Ati? I do not think so "

When Nvidia improves the drivers, and by a good margin, the GF-FX will be ahead in every benchmark. So I have to ask, what do you mean by "will it be enough to beat Ati? I do not think so " :wtf: He said beat ATI, not beat 9700 Pro. If they take the same time frame as they did to release the detXP's for the GF3's, ATI will have two cards newer than the 9700 Pro on the market. Will they improve their drivers by a good margin...I can almost guarantee it...can they do it fast enough to counter the R350 and possibly the R400...almost no way for the r350, and debatable for the R400.

Filburt
01-31-03, 08:25 PM
I think it's doubtful they will beat even the 9700 Pro in bandwidth intensive things such as using FSAA + Anisotropic filtering at fairly high resolution when both speed and quality are considered. They can up the speed all the want by doing the same blurvision trick they've been doing with 2x and QC...but that doesn't make it any better. If I were an nvidia-only buying person, I would wait for the NV35. The reason being because by then they will have already gotten the introduction of an array of features over with...and can then concentrate on making them run well.

Now, since I'm personally willing to buy from either company...I'm a bit more optimistic about the R400 than the NV35. Reason being it's had plenty of time to be developed...and it seems as if ATi's ability has vastly improved these days. However if the NV35 proves to be a better value, I'll buy that...I don't really much care. If the NV40 is slated for very soon after the R400, it's likely I'll buy that instead unless it is too expensive or is too flawed. Same goes for the ATi cards I suppose.

Steppy
01-31-03, 09:14 PM
If you follow NV's roadmap for the past few years you can pretty well guestimate when the new GPU's will be available.

NV10 Fall of 99
NV15(GF2) Spring 2000 refresh 6 months later(Pro and Ultra)
NV20(GF3) Spring 2001 refresh 6 months later(Ti's)
NV25(GF4) Spring 2002 refresh 6 months later(8x versions)
NV30(GFFX) Fall 2002 refresh 6 months later(Ultra)
The specs/hardware(500Mhz core, and DDR-II at 500 wasn't available) of the ultra would have been impossible back in fall, so it likely would have followed the previous path and only the non-ultra would have come out back in fall(whose hardware was available)the missed lauch basically just made the original AND the refresh debut when only the refresh normally would have.
NV35(GFFX2) Fall 2003 refresh 6 months later
NV40(GFFX3) Spring 2004 if they were able to push it to 6 months between cores like the NV30 from NV25 was(unsuccessfully) Fall 2004 if it is the typical year between cores(which NV25-30 became). If NV35 follows the typical timeframe to, you're looking at Spring 2004 for NV35, and spring 2005 for NV40. NV is in a precarious situation unless they CAN ACTUALLY regain 6 months on ATI.

Steppy
01-31-03, 09:23 PM
Anyway, since the poll asks 2 different questions should the answers have 2 answers also ie

1. yes,no
2. yes,yes,
3. no,no
4. no,yes

There really aren't options to agree with one part and not with the other

Kruno
01-31-03, 09:39 PM
I doubt the NV30 will ever be able to perform as good as the R300 with max IQ settings. (8x FSAA, 64-tap aniso etc... at 1280x960)

I know I play current games at 128 tap aniso, 6xFSAA and 1280x860. :)
All run perfect.

The Baron
01-31-03, 09:54 PM
If ATI follows a 6 month product cycle, and R350 is released in March (as a lot of the rumors say), then R400 CANNOT be released in July.

PERIOD END OF STORY. Maybe a PAPER launch in July, but you could see GFFX style delays on that one. Wouldn't surprise me either way--ATI now has to go back to a continual 6 month cycle after skipping one after 8500.

Yes, I'm pissy because I have a fever, am really tired, and am really really sick of hearing about how fast ATI is going to put cards on the market. :p

Steppy
01-31-03, 10:00 PM
Originally posted by The Baron
If ATI follows a 6 month product cycle, and R350 is released in March (as a lot of the rumors say), then R400 CANNOT be released in July.

PERIOD END OF STORY. Maybe a PAPER launch in July, but you could see GFFX style delays on that one. Wouldn't surprise me either way--ATI now has to go back to a continual 6 month cycle after skipping one after 8500.

Yes, I'm pissy because I have a fever, am really tired, and am really really sick of hearing about how fast ATI is going to put cards on the market. :p
R100 Fall 2000
R200 Fall 2001
R300 Fall 2002
R400 <Hmmm...wonder when it should hit the streets, and has already been hinted at. It may be LAUNCHED in LATE July, putting cards out in late August to early September almost exactly 12 months after the R300.

StealthHawk
01-31-03, 11:15 PM
IMO the gf4 never got incredible boosts in speed.

you can count the AF incident...with one caveat. we all know, or should know that the AF was subpar in both OGL and D3D, with AF having a bigger performance hit on the gf4 than the gf3. nvidia optimized AF in OGL starting with the 28.90 drivers. so in my mind performance was placed where it should have been in the beginning. now, comparing the scores of the newest drivers to the 28.90 drivers might make a case. but then the driver improvement becomes a lot smaller.

tb77
01-31-03, 11:42 PM
Performance will improve. Nice article, but I would like to see the performance improvements of an directx game/benchmark.

"Quake 3 Performance - 1600x1200"

wrong picture link?

Thomas

The Baron
01-31-03, 11:52 PM
Stealth, GF4 never got the boost in speed because it was already pretty well optimized (revision of the GF3 core and all that).

And Thomas, yes, that's been pointed out, it should be fixed soon.

I'm going to go back to slipping in and out of consciousness now. :p

hithere
02-01-03, 12:13 AM
Has anyone considered that an IQ boost to the current state of anti-aliasing on the FX might make it perform WORSE than it already does? Maybe that is why the IQ is low in the first place?

Steppy
02-01-03, 12:18 AM
The fact that 2xAA and Quincuxx got labeled together and both have problems with screenshots because a "post frame buffer" filter hasn't been applied yet leads me to beleive that the only think missing from the screenshots is the "blur" filter quincuxx uses. My guess is that 2xaa is just a better blur now. Either way, I doubt the 2xaa is gonna approach the quality of the 9700 AA at equal settings.

MUYA
02-01-03, 02:59 AM
If you guys look sat this thread with benches of dx9 demos (albeit from ATI..still dx9 nontheless) here and pointed out in an another thread here too;

http://www.beyond3d.com/forum/viewtopic.php?t=4041&postdays=0&postorder=asc&start=0



You will see the FX getting beaten the crap out and even a ATI9000 manages to stay with the FX in dx8, dx9 tests. The 9700pro is around 100% or more faster on dx9 ps2 tests etc.

To me, this seems as if that the drivers are really in a bad state to be on par with 2pipeline powered product(ati9000). I cannot believe for one minute that nv would engineerand release such a piece of bad silicon and hope the community will not figure that out. The community is ingineous and will finds way to test and counter claim every claim each manufacturer make! But in my mind there is (I hope) more room for driver optimization etc etc because we are seing benches where the FX 's results is comparable to that of a 2 pipeline board and well it doesn't seem fathomable that nV would in their right mind release such a hardware that would be whooped so bad by an existing product or be shown by a 2pipeline 9000pro and still expect ppl to pay $399 for it.
Yes they had a huge amount of time to get the drivers right (since Comdex) but then again debugging the crap out of drivers for a new hardware with new a radically new architecture will probably take much more time! All marks and credit to ATI though for catching nV with their pants down and they have made nV nervous enough to actually proabbly(speculation) dictate the price of a FX ultra from estimated $500 to a now MSRP $399! But I still think there much more space for improvement in certain departments through drivers for the FX for it to cath up in them dx9n dx8 benchmarks. I do doubt that maybe their claim for FSAA for free is probably all hype than substance though.

MUYA

Kruno
02-01-03, 03:03 AM
Even if nVidia "fix" everything in future, it doesn't change the fact that people will still be bashing it from it's first reviews.

StealthHawk
02-01-03, 05:21 AM
Originally posted by The Baron
Stealth, GF4 never got the boost in speed because it was already pretty well optimized (revision of the GF3 core and all that).

yes i know that. the point is that i don't believe you can come to the conclusion that "driver optimizations can make great performance increases. look at the gf4 AF settings, they showed incredible gains" and infer that gfFX will also have similar large percentage gains. and this simply because AF was horribly unoptimized from the start on the gf4.

i'm rather mixed in my opinion of the state of gfFX's drivers. surely with all the delays they've had a lot of time to eek out more performance from the card than we would have seen had the card shipped 1 month earlier or even a few months earlier.

that being said, i don't doubt we won't get drivers later that improve performance by as much as 20% within a few months. in a few situations performance will be higher, in others less. butn i don't think there will be a magic driver revision that pumps up high image quality scores by 40%

Uttar
02-01-03, 09:20 AM
The GFFX could easily beat the Radeon 9700 Pro in most cases by about 15% or 30% in best-case scenarios with better drivers ( which would release in the next two months )

The problem is that this won't be sufficent to compete IQ-wise with the R300, and won't be sufficent performance-wise against the R350.
So the GFFX selling point really becomes features...


Uttar

Steppy
02-01-03, 10:04 AM
and it has no compelling features over either the R300 or R350 to sell it. The only thing I see it having over the 9700 is the ps 2.0+ and vertex shader x.x+. Since we're at least a year away from games that just use baseline DX9 compliance, those "+'s" will and do mean absolutely nothing.

The Baron
02-01-03, 11:36 AM
yes i know that. the point is that i don't believe you can come to the conclusion that "driver optimizations can make great performance increases. look at the gf4 AF settings, they showed incredible gains" and infer that gfFX will also have similar large percentage gains. and this simply because AF was horribly unoptimized from the start on the gf4.

That was just because neither Mike nor me had an original GeForce around. I might have to scrounge up some of those 11.xx GF3 drivers, install them (although I doubt they'd install on XP), and compare them with the 4x.xx series... :p